[openib-general] Re: [PATCH] cm refcount race fix

[Roland Dreier]

    Michael> We'll have to change each and every access to the locked
    Michael> version then though.  It would be a big risky change. How
    Michael> about the following compromize?

I have to disagree.  I don't think it's big or risky at all... there's
only 11 uses of refcount in cm.c, and one of them is initialization.
If we can't get that right then we should just give up.

 - R.